Setup: alerts, AI key and your own copy

Where the settings are

Open the repository on GitHub → Settings → Secrets and variables → Actions. There are two tabs:

• Secrets: hidden values (tokens, keys). Nobody can read them back, not even you, and they are masked in logs.
• Variables: ordinary values that are not secret (a language code, a model name).

A change applies from the next run. To run now: Actions → research-agents → Run workflow.

Telegram alerts

The agents can message you when something important happens: the market regime changes, the H-020 rule switches in or out, a hypothesis gets a new verdict, a data source breaks, and once a week a digest.

1. In Telegram open @BotFather. Use an existing bot (/mybots → your bot → API Token) or create one with /newbot.
2. Open your bot and press Start, otherwise it is not allowed to write to you.
3. Find your chat ID: write to @userinfobot, it replies with a number like 123456789. For a group or a channel add the bot there; their IDs start with -100.
4. In Settings → Secrets and variables → Actions → Secrets → New repository secret create TELEGRAM_BOT_TOKEN (the token) and TELEGRAM_CHAT_ID (the ID).
5. Language: in Variables → New repository variable create TELEGRAM_LANG with en, ru, es, pt, tr, vi or zh. Without it alerts are in English. Every alert ends with a link to the dashboard in the same language.
6. Test: Actions → research-agents → Run workflow, tick test_alert, run. You should get "Telegram alerts are connected" within a few minutes. If the step fails, the token or ID is wrong, or the bot was never started.

Alerts are sent only when something changes, so quiet days are silent.

AI model key (optional)

All tests are done by the engine with fixed rules; a model is never asked whether an idea works. A model key adds two helpers:

• Weekly digest in plain words. Without a key you get the raw weekly reports in English. With a key a model rewrites them into a short briefing in the language of TELEGRAM_LANG, using only numbers from the reports and never suggesting trades.
• New hypothesis drafts once a month (the 1st). The model reads what has been tested and proposes new ideas as a pull request with status draft. You decide what to register; nothing is tested until you do.

Free option: Groq. Sign up at console.groq.com → API Keys → Create API Key, and save it as the secret LLM_API_KEY. No payment needed (the free tier has daily limits, plenty for a weekly digest). The model is chosen automatically from the free models Groq currently offers (they get renamed and retired); set CRL_LLM_MODEL to fix one. Any other OpenAI-compatible service (OpenRouter, Gemini) works too: put its address in the variable CRL_LLM_BASE_URL. If both keys are set, LLM_API_KEY is used.

How to connect Anthropic (paid):

1. Go to console.anthropic.com, sign up and add credit under Billing. The API is paid per use and separately from a Claude subscription. A weekly digest and a monthly proposal cost cents a month. Check first that your country is on Anthropic's list of supported regions.
2. API Keys → Create Key, copy the key (sk-ant-…). It is shown only once.
3. In Settings → Secrets and variables → Actions → Secrets create ANTHROPIC_API_KEY with the key.
4. Optional: variable CRL_LLM_MODEL to choose another model (default claude-sonnet-5-5).
5. For the monthly drafts to arrive as pull requests: Settings → Actions → General → Allow GitHub Actions to create and approve pull requests.

To switch it off, delete the secret. Revoke a leaked key in the console at once.

Run your own copy

Everything runs on GitHub's free servers; nothing is needed on your computer.

1. Open https://github.com/f1devbin/Crypto-Research-Labs and press Fork.
2. In your fork: Settings → Pages → Source: GitHub Actions.
3. Actions → enable workflows if asked → research-agents → Run workflow. The first run takes about 20 minutes (it downloads the price history). Then it runs by itself every day.
4. Your dashboard appears at https://<your-name>.github.io/Crypto-Research-Labs/.
5. Add Telegram and the AI key as described above if you want them.

To add your own idea, see CONTRIBUTING.md or post it in Discussions → Ideas.

All settings at a glance